Codestral 22B
Mistral's 22B fluent-code model — 80+ languages under the non-production MNPL license.
About
Codestral-22B (May 2024) is Mistral AI's code model: fluent in 80+ programming languages, strong on code generation and fill-in-the-middle, in a 22B package.
Licensed under the Mistral Non-Production License (MNPL): custom, NOT open source — no production use. This is the critical caveat: the weights are research/test-only. (Contrast Devstral Small's Apache 2.0.) 20k Hugging Face downloads. A 2024 model superseded by newer options, and the MNPL rules out products entirely. Useful for research and evaluation; for production Mistral code models, use the API or Devstral.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-claimed at release (2024): strong HumanEval for 22B — treat as provider-claimed and dated. No current leaderboard presence.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 5/10
Scored 5/10 (metrics tier). Capable 22B multilingual code model with real adoption (20k downloads), but the MNPL is disqualifying for products: (1) no production use — research/test only; (2) 2024-era, superseded; (3) benchmark claims are dated provider reports; (4) Mistral's own Devstral line offers Apache-2.0 alternatives. Research and evaluation use only.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.6/10
- Maintenance · recency, releases, issue load
- 4.0/10
- Openness · license + docs
- 3.2/10
- Manual review · sanity read + red-flag check
- 5.0/10
- · HF downloads 19,952 (peer-normalized)
- · released 867 days ago
- · manual review: capable 22B but MNPL disqualifies products; 2024-era; Devstral offers Apache-2.0 alternatives; license: Mistral Non-Production License (MNPL) — NO production use; research/test only